Video wird geladen...

Video konnte nicht geladen werden

Beim Laden dieses Videos ist ein Problem aufgetreten. Dies könnte an einem vorübergehenden Netzwerkproblem liegen oder das Video ist möglicherweise nicht verfügbar.

Novel view synthesis has long been a core challenge in 3D vision. But how much 3D inductive bias is truly needed? —Surprisingly, very little! Introducing "LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias"—a fully transformer-based approach that enables scalable, generalizable, and fully data-driven novel view synthesis,... show more

Haian Jin

1,244 subscribers

114,947 Aufrufe • vor 1 Jahr •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

26 Kommentare

Profilbild von Haian Jin

Haian Jinvor 1 Jahr

(2/6) We introduce two architectures: (1) an encoder-decoder LVSM, and (2) a decoder-only LVSM. Both models bypass the 3D inductive biases used in previous methods—from 3D representations (e.g., NeRF, 3DGS) to network designs (e.g., epipolar projections, plane sweeps)—addressing novel view synthesis with a fully data-driven approach.

Profilbild von Haian Jin

Haian Jinvor 1 Jahr

(3/6) Our methods can work for both objects and scenes from sparse inputs (2-4 views), and our best models outperform previous state-of-the-art methods by 1.5 to 3.5 dB PSNR.

Profilbild von Haian Jin

Haian Jinvor 1 Jahr

(4/6) We observe that our LVSM also works with a single input view for many cases, despite only being trained with multi-view inputs. This observation shows the capability of LVSM to understand the 3D world, e.g. understanding depth, rather than just performing pixel-level view interpolation.

Profilbild von Haian Jin

Haian Jinvor 1 Jahr

(5/6) Notably, our models surpass all previous methods even with reduced computational resources (1-2 GPUs).

Profilbild von Haian Jin

Haian Jinvor 1 Jahr

(6/6) Thanks for my wonderful collaborators's support and help! @hanwenjiang1 @HaoTan5 @KaiZhang9546 @Sai__Bi @tianyuanzhang99 @fujun_luan @Jimantha @zexiangxu

Profilbild von Junlin Han

Junlin Hanvor 1 Jahr

very surprised with the extremely good lpips results, congrats on making a huge move in fast 3d reconstruction!

Profilbild von Haian Jin

Haian Jinvor 1 Jahr

Thank you, Junlin!

Profilbild von Springblade 🇺🇸

Springblade 🇺🇸vor 1 Jahr

This is super cool, I've been waiting for this to be invented! For a few years now, I've been taking certain photos with a couple of additional shots from different angles, knowing that some day I would be able to make 3D views! The world is catching up to me! Cool!

Profilbild von Sida Peng

Sida Pengvor 1 Jahr

Very cool!

Profilbild von Haian Jin

Haian Jinvor 1 Jahr

Thank you, Sida!

Profilbild von Matthias Niessner

Matthias Niessnervor 1 Jahr

Pretty cool - congrats!!!

Profilbild von Haian Jin

Haian Jinvor 1 Jahr

Thank you!

Profilbild von Robert Scoble

Robert Scoblevor 1 Jahr

I can’t wait to play in your Holodeck.

Profilbild von Springblade 🇺🇸

Springblade 🇺🇸vor 1 Jahr

This is super cool!

Profilbild von wang zy

wang zyvor 1 Jahr

It is really a great work! I have some question about it. 1. How much GPU memory is required for inference? 2. What is the inference speed? Is it faster than stable video diffusion?

Profilbild von Deptho.ai

Deptho.aivor 1 Jahr

So powerful! 🔥

Profilbild von chener

chenervor 1 Jahr

What do you need? Just a few pictures or pictures with accurate camera transformations?

Profilbild von Ronal Dragon

Ronal Dragonvor 1 Jahr

what about dust3r?

Profilbild von Season

Seasonvor 1 Jahr

Great work! How is the visual quality so good without using diffusion models? Especially when you zoom into details.

Profilbild von Abdullah Hamdi

Abdullah Hamdivor 1 Jahr

Cool work !

Profilbild von Daniel

Danielvor 1 Jahr

🔥🔥🚀

Profilbild von daniel

danielvor 1 Jahr

Do you think 3D Inductive Biases are unnecessary (in the context of NVS) Or do you think we don’t have enough 3D data / havent found a good representation yet?

Profilbild von nama

namavor 1 Jahr

wow!! really cool.. how sensitive is the method to having noisy poses?

Profilbild von Julien

Julienvor 1 Jahr

Congrats! I like the capacity of the model to have good results with Single Image even if the model is trained for Multiple. And any info about the inference speed and code release date?

Profilbild von JB

JBvor 1 Jahr

Wen app?

Profilbild von Zaky Vids

Zaky Vidsvor 1 Jahr

@theworldlabs Since many of us are waiting for so long and most of us are not really going to get access anytime soon you guys might as well look into this approach too

Ähnliche Videos

🍺 LagerNVS (CVPR 2026) 🍺 LagerNVS is a generalizable, feed-forward, real-time Novel View Synthesis network which - performs rendering in real time, - generalizes to in-the-wild data, - works with and without known source cameras, - sets a new state-of-the-art among deterministic methods, - can be paired with a diffusion decoder for generative extrapolation. LagerNVS shows that 3D biases are useful for Novel View Synthesis but explicit 3D representations are not required to achieve them. We use 3D biases in (1) architecture design and (2) pre-training: (1) In NVS with explicit 3D representations (3DGS, NeRF) reconstruction is typically difficult and slow, but rendering is much faster and simpler. We mimic this process in the network design: we use a large (1B params) encoder and a small, lightweight decoder (ViT-B). This allows increasing the network capacity while still achieving real-time rendering. (2) The encoder, initialized from VGGT, was pre-trained with 3D reconstruction objectives, making the initial features 3D aware. Both substantially improve performance. Project page: Code: Paper: Models: Work done with Jianyuan Minghao Chen Christian Rupprecht and Andrea Vedaldi

🍺 LagerNVS (CVPR 2026) 🍺 LagerNVS is a generalizable, feed-forward, real-time Novel View Synthesis network which - performs rendering in real time, - generalizes to in-the-wild data, - works with and without known source cameras, - sets a new state-of-the-art among deterministic methods, - can be paired with a diffusion decoder for generative extrapolation. LagerNVS shows that 3D biases are useful for Novel View Synthesis but explicit 3D representations are not required to achieve them. We use 3D biases in (1) architecture design and (2) pre-training: (1) In NVS with explicit 3D representations (3DGS, NeRF) reconstruction is typically difficult and slow, but rendering is much faster and simpler. We mimic this process in the network design: we use a large (1B params) encoder and a small, lightweight decoder (ViT-B). This allows increasing the network capacity while still achieving real-time rendering. (2) The encoder, initialized from VGGT, was pre-trained with 3D reconstruction objectives, making the initial features 3D aware. Both substantially improve performance. Project page: Code: Paper: Models: Work done with Jianyuan Minghao Chen Christian Rupprecht and Andrea Vedaldi

Stan Szymanowicz

32,815 Aufrufe • vor 6 Monaten

Large-scale 3D Scene Generation (all scenes are real-time rendered)!! Physically-grounded generative data without hallucinations is the missing link for robot learning and testing at scale. We introduce a method that directly generates large-scale 3D driving scenes with accurate geometry, allowing for causal view synthesis and generation with object permanence and explicit 3D geometry. This also allows for extreme trajectory extrapolation without failure! We also show that we can build fully data-driven simulators for end-to-end learning with this approach. Project: with the amazing team of Julian Ost, Amogh Joshi , Andrea Ramazzina, Maximilian Bömer, Mario Bijelic.

Large-scale 3D Scene Generation (all scenes are real-time rendered)!! Physically-grounded generative data without hallucinations is the missing link for robot learning and testing at scale. We introduce a method that directly generates large-scale 3D driving scenes with accurate geometry, allowing for causal view synthesis and generation with object permanence and explicit 3D geometry. This also allows for extreme trajectory extrapolation without failure! We also show that we can build fully data-driven simulators for end-to-end learning with this approach. Project: with the amazing team of Julian Ost, Amogh Joshi , Andrea Ramazzina, Maximilian Bömer, Mario Bijelic.

Felix Heide

27,833 Aufrufe • vor 1 Jahr

Robot Learning needs 4D world models! Robot Learning needs 4D world models! Robot Learning needs 4D world models! We introduce TesserAct, a 4D embodied world model that can simulate how agents interact with the 3D world over time! We achieve this by simply extending a pre-trained 2D video generation model to jointly predict RGB, depth, and surface normals. It enables: 1️⃣ Much better policy learning in the wild 2️⃣ Temporal + spatial coherence in 4D dynamic prediction 3️⃣ Novel view synthesis for embodied scenes Code: Paper Link: Project page:

Robot Learning needs 4D world models! Robot Learning needs 4D world models! Robot Learning needs 4D world models! We introduce TesserAct, a 4D embodied world model that can simulate how agents interact with the 3D world over time! We achieve this by simply extending a pre-trained 2D video generation model to jointly predict RGB, depth, and surface normals. It enables: 1️⃣ Much better policy learning in the wild 2️⃣ Temporal + spatial coherence in 4D dynamic prediction 3️⃣ Novel view synthesis for embodied scenes Code: Paper Link: Project page:

Chuang Gan

43,265 Aufrufe • vor 1 Jahr

WorldExplorer: Towards Generating Fully Navigable 3D Scenes Contributions: • We introduce the first method for generating 3D scenes from text that supports high-quality view synthesis while enabling exploration across a wide range of camera poses. • We propose an iterative scene expansion strategy using video diffusion models, driven by trajectory sampling and adaptive collision detection. • We design a scene memory mechanism that conditions each video generation step on relevant past frames, improving view consistency and overall scene coherence.

WorldExplorer: Towards Generating Fully Navigable 3D Scenes Contributions: • We introduce the first method for generating 3D scenes from text that supports high-quality view synthesis while enabling exploration across a wide range of camera poses. • We propose an iterative scene expansion strategy using video diffusion models, driven by trajectory sampling and adaptive collision detection. • We design a scene memory mechanism that conditions each video generation step on relevant past frames, improving view consistency and overall scene coherence.

MrNeRF

23,814 Aufrufe • vor 1 Jahr

DL3DV-10K: A Large-Scale Scene Dataset for Deep Learning-based 3D Vision paper page: We have witnessed significant progress in deep learning-based 3D vision, ranging from neural radiance field (NeRF) based 3D representation learning to applications in novel view synthesis (NVS). However, existing scene-level datasets for deep learning-based 3D vision, limited to either synthetic environments or a narrow selection of real-world scenes, are quite insufficient. This insufficiency not only hinders a comprehensive benchmark of existing methods but also caps what could be explored in deep learning-based 3D analysis. To address this critical gap, we present DL3DV-10K, a large-scale scene dataset, featuring 51.2 million frames from 10,510 videos captured from 65 types of point-of-interest (POI) locations, covering both bounded and unbounded scenes, with different levels of reflection, transparency, and lighting. We conducted a comprehensive benchmark of recent NVS methods on DL3DV-10K, which revealed valuable insights for future research in NVS. In addition, we have obtained encouraging results in a pilot study to learn generalizable NeRF from DL3DV-10K, which manifests the necessity of a large-scale scene-level dataset to forge a path toward a foundation model for learning 3D representation.

DL3DV-10K: A Large-Scale Scene Dataset for Deep Learning-based 3D Vision paper page: We have witnessed significant progress in deep learning-based 3D vision, ranging from neural radiance field (NeRF) based 3D representation learning to applications in novel view synthesis (NVS). However, existing scene-level datasets for deep learning-based 3D vision, limited to either synthetic environments or a narrow selection of real-world scenes, are quite insufficient. This insufficiency not only hinders a comprehensive benchmark of existing methods but also caps what could be explored in deep learning-based 3D analysis. To address this critical gap, we present DL3DV-10K, a large-scale scene dataset, featuring 51.2 million frames from 10,510 videos captured from 65 types of point-of-interest (POI) locations, covering both bounded and unbounded scenes, with different levels of reflection, transparency, and lighting. We conducted a comprehensive benchmark of recent NVS methods on DL3DV-10K, which revealed valuable insights for future research in NVS. In addition, we have obtained encouraging results in a pilot study to learn generalizable NeRF from DL3DV-10K, which manifests the necessity of a large-scale scene-level dataset to forge a path toward a foundation model for learning 3D representation.

AK

49,929 Aufrufe • vor 2 Jahren

V3D Video Diffusion Models are Effective 3D Generators Automatic 3D generation has recently attracted widespread attention. Recent methods have greatly accelerated the generation speed, but usually produce less-detailed objects due to limited model capacity or 3D data. Motivated by recent advancements in video diffusion models, we introduce V3D, which leverages the world simulation capacity of pre-trained video diffusion models to facilitate 3D generation. To fully unleash the potential of video diffusion to perceive the 3D world, we further introduce geometrical consistency prior and extend the video diffusion model to a multi-view consistent 3D generator. Benefiting from this, the state-of-the-art video diffusion model could be fine-tuned to generate 360degree orbit frames surrounding an object given a single image. With our tailored reconstruction pipelines, we can generate high-quality meshes or 3D Gaussians within 3 minutes. Furthermore, our method can be extended to scene-level novel view synthesis, achieving precise control over the camera path with sparse input views. Extensive experiments demonstrate the superior performance of the proposed approach, especially in terms of generation quality and multi-view consistency

V3D Video Diffusion Models are Effective 3D Generators Automatic 3D generation has recently attracted widespread attention. Recent methods have greatly accelerated the generation speed, but usually produce less-detailed objects due to limited model capacity or 3D data. Motivated by recent advancements in video diffusion models, we introduce V3D, which leverages the world simulation capacity of pre-trained video diffusion models to facilitate 3D generation. To fully unleash the potential of video diffusion to perceive the 3D world, we further introduce geometrical consistency prior and extend the video diffusion model to a multi-view consistent 3D generator. Benefiting from this, the state-of-the-art video diffusion model could be fine-tuned to generate 360degree orbit frames surrounding an object given a single image. With our tailored reconstruction pipelines, we can generate high-quality meshes or 3D Gaussians within 3 minutes. Furthermore, our method can be extended to scene-level novel view synthesis, achieving precise control over the camera path with sparse input views. Extensive experiments demonstrate the superior performance of the proposed approach, especially in terms of generation quality and multi-view consistency

AK

31,997 Aufrufe • vor 2 Jahren

Alibaba presents MIMO Controllable Character Video Synthesis with Spatial Decomposed Modeling Character video synthesis aims to produce realistic videos of animatable characters within lifelike scenes. As a fundamental problem in the computer vision and graphics community, 3D works typically require multi-view captures for per-case training, which severely limits their applicability of modeling arbitrary characters in a short time. Recent 2D methods break this limitation via pre-trained diffusion models, but they struggle for pose generality and scene interaction. To this end, we propose MIMO, a novel framework which can not only synthesize character videos with controllable attributes (i.e., character, motion and scene) provided by simple user inputs, but also simultaneously achieve advanced scalability to arbitrary characters, generality to novel 3D motions, and applicability to interactive real-world scenes in a unified framework. The core idea is to encode the 2D video to compact spatial codes, considering the inherent 3D nature of video occurrence. Concretely, we lift the 2D frame pixels into 3D using monocular depth estimators, and decompose the video clip to three spatial components (i.e., main human, underlying scene, and floating occlusion) in hierarchical layers based on the 3D depth. These components are further encoded to canonical identity code, structured motion code and full scene code, which are utilized as control signals of synthesis process. The design of spatial decomposed modeling enables flexible user control, complex motion expression, as well as 3D-aware synthesis for scene interactions. Experimental results demonstrate effectiveness and robustness of the proposed method.

Alibaba presents MIMO Controllable Character Video Synthesis with Spatial Decomposed Modeling Character video synthesis aims to produce realistic videos of animatable characters within lifelike scenes. As a fundamental problem in the computer vision and graphics community, 3D works typically require multi-view captures for per-case training, which severely limits their applicability of modeling arbitrary characters in a short time. Recent 2D methods break this limitation via pre-trained diffusion models, but they struggle for pose generality and scene interaction. To this end, we propose MIMO, a novel framework which can not only synthesize character videos with controllable attributes (i.e., character, motion and scene) provided by simple user inputs, but also simultaneously achieve advanced scalability to arbitrary characters, generality to novel 3D motions, and applicability to interactive real-world scenes in a unified framework. The core idea is to encode the 2D video to compact spatial codes, considering the inherent 3D nature of video occurrence. Concretely, we lift the 2D frame pixels into 3D using monocular depth estimators, and decompose the video clip to three spatial components (i.e., main human, underlying scene, and floating occlusion) in hierarchical layers based on the 3D depth. These components are further encoded to canonical identity code, structured motion code and full scene code, which are utilized as control signals of synthesis process. The design of spatial decomposed modeling enables flexible user control, complex motion expression, as well as 3D-aware synthesis for scene interactions. Experimental results demonstrate effectiveness and robustness of the proposed method.

AK

149,079 Aufrufe • vor 2 Jahren

Fast View Synthesis of Casual Videos paper page: Novel view synthesis from an in-the-wild video is difficult due to challenges like scene dynamics and lack of parallax. While existing methods have shown promising results with implicit neural radiance fields, they are slow to train and render. This paper revisits explicit video representations to synthesize high-quality novel views from a monocular video efficiently. We treat static and dynamic video content separately. Specifically, we build a global static scene model using an extended plane-based scene representation to synthesize temporally coherent novel video. Our plane-based scene representation is augmented with spherical harmonics and displacement maps to capture view-dependent effects and model non-planar complex surface geometry. We opt to represent the dynamic content as per-frame point clouds for efficiency. While such representations are inconsistency-prone, minor temporal inconsistencies are perceptually masked due to motion. We develop a method to quickly estimate such a hybrid video representation and render novel views in real time. Our experiments show that our method can render high-quality novel views from an in-the-wild video with comparable quality to state-of-the-art methods while being 100x faster in training and enabling real-time rendering.

Fast View Synthesis of Casual Videos paper page: Novel view synthesis from an in-the-wild video is difficult due to challenges like scene dynamics and lack of parallax. While existing methods have shown promising results with implicit neural radiance fields, they are slow to train and render. This paper revisits explicit video representations to synthesize high-quality novel views from a monocular video efficiently. We treat static and dynamic video content separately. Specifically, we build a global static scene model using an extended plane-based scene representation to synthesize temporally coherent novel video. Our plane-based scene representation is augmented with spherical harmonics and displacement maps to capture view-dependent effects and model non-planar complex surface geometry. We opt to represent the dynamic content as per-frame point clouds for efficiency. While such representations are inconsistency-prone, minor temporal inconsistencies are perceptually masked due to motion. We develop a method to quickly estimate such a hybrid video representation and render novel views in real time. Our experiments show that our method can render high-quality novel views from an in-the-wild video with comparable quality to state-of-the-art methods while being 100x faster in training and enabling real-time rendering.

AK

20,668 Aufrufe • vor 2 Jahren

[SIGGRAPH Asia '24 (TOG)] Representing Long Volumetric Video with Temporal Gaussian Hierarchy Contributions: • We introduce a novel, efficient, and expressive Temporal Gaussian Hierarchy representation for long volumetric video. To our knowledge, our method is the first approach capable of handling minutes of volumetric video data. • We propose a Compact Appearance Model and a new rasterization implementation to facilitate real-time, high-quality dynamic view synthesis while maintaining a compact size. • We propose a system to efficiently model long volumetric videos for the first time and demonstrate state-of-the-art dynamic view synthesis quality on the Neural3DV [Li et al. 2022], ENeRF-Outdoor [Lin et al. 2022], and MobileStage [Xu et al. 2024b] datasets, while also achieving the best rendering speed with reduced training cost and memory usage.

[SIGGRAPH Asia '24 (TOG)] Representing Long Volumetric Video with Temporal Gaussian Hierarchy Contributions: • We introduce a novel, efficient, and expressive Temporal Gaussian Hierarchy representation for long volumetric video. To our knowledge, our method is the first approach capable of handling minutes of volumetric video data. • We propose a Compact Appearance Model and a new rasterization implementation to facilitate real-time, high-quality dynamic view synthesis while maintaining a compact size. • We propose a system to efficiently model long volumetric videos for the first time and demonstrate state-of-the-art dynamic view synthesis quality on the Neural3DV [Li et al. 2022], ENeRF-Outdoor [Lin et al. 2022], and MobileStage [Xu et al. 2024b] datasets, while also achieving the best rendering speed with reduced training cost and memory usage.

MrNeRF

79,493 Aufrufe • vor 1 Jahr

New paper from Apple - Sharp Monocular View Synthesis in Less than a Second Mescheder et al. @ Apple just released a very impressive paper (congrats! 🎉🥳). You give it an image and it generates a really great looking 3d Gaussian representation. Uses depth pro. It's really good. The model is about 3GB, and takes ~5-10s on my M1 Max per image. Single feedforward pass network. The video is using "Metal Splatter" to view the .ply from ml-sharp on Apple Vision Pro. Some really wow moments today trying it on different scenes.

New paper from Apple - Sharp Monocular View Synthesis in Less than a Second Mescheder et al. @ Apple just released a very impressive paper (congrats! 🎉🥳). You give it an image and it generates a really great looking 3d Gaussian representation. Uses depth pro. It's really good. The model is about 3GB, and takes ~5-10s on my M1 Max per image. Single feedforward pass network. The video is using "Metal Splatter" to view the .ply from ml-sharp on Apple Vision Pro. Some really wow moments today trying it on different scenes.

Tim Davison ᯅ

133,059 Aufrufe • vor 9 Monaten

Want to create an avatar from a single image? FlexAvatar is a transformer model that creates full 360°, high-quality, and expressive 3D head avatar from just a single portrait image in minutes. Real-time Demo: FlexAvatar's lightweight architecture allows both animation and rendering in real-time, enabling interactive user experiences. To create a new 3D head avatar, only one image is required, e.g., from a webcam. The final avatar is ready after 2 minutes. Architecture: Under the hood, FlexAvatar adopts a transformer-based encoder-decoder design. The encoder maps the input image onto a latent avatar space, while the decoder produces 3D Gaussian attribute maps by incorporating the animation signal via cross-attention. The model learns all facial animations directly from the data without relying on pre-built 3D face models. This equips the avatars with realistic facial expressions. The internal avatar latent space can be conveniently used to integrate additional observations of a person via fitting. This enables use-cases where more than one image of a person is available, e.g., from a phone scan of the person. We train jointly on 2D monocular videos and multi-view data. However, in monocular videos, the animation signal leaks the target viewpoint, causing the model to produce incomplete 3D heads. We call this phenomenon entanglement of driving signal and target viewpoint. To prevent entanglement, we introduce bias sinks. These are learnable tokens that indicate whether a training sample stems from a monocular or a multi-view dataset. During training, the model learns to produce incomplete 3D heads only when the monocular token is present. During inference, FlexAvatar then always uses the multi-view token for which the model has learned to produce complete 3D heads. This simple design allows to combine the generalizability from monocular data with the quality of multi-view data. FlexAvatar summary: - Input: Single-image, phone scan, or monocular video - Output: Full 360° head avatar - Expressive animations - Real-time rendering and animation - Generalization to any portrait - Create a new avatar in 2 minutes - Use bias sinks to combine 2D and 3D data 🏠 🌍 🎥 Great work by Tobias Kirschstein and Simon Giebenhain!

Matthias Niessner

96,371 Aufrufe • vor 9 Monaten

Text-to-3D Generation with Bidirectional Diffusion using both 2D and 3D priors paper page: Most 3D generation research focuses on up-projecting 2D foundation models into the 3D space, either by minimizing 2D Score Distillation Sampling (SDS) loss or fine-tuning on multi-view datasets. Without explicit 3D priors, these methods often lead to geometric anomalies and multi-view inconsistency. Recently, researchers have attempted to improve the genuineness of 3D objects by directly training on 3D datasets, albeit at the cost of low-quality texture generation due to the limited texture diversity in 3D datasets. To harness the advantages of both approaches, we propose Bidirectional Diffusion(BiDiff), a unified framework that incorporates both a 3D and a 2D diffusion process, to preserve both 3D fidelity and 2D texture richness, respectively. Moreover, as a simple combination may yield inconsistent generation results, we further bridge them with novel bidirectional guidance. In addition, our method can be used as an initialization of optimization-based models to further improve the quality of 3D model and efficiency of optimization, reducing the generation process from 3.4 hours to 20 minutes. Experimental results have shown that our model achieves high-quality, diverse, and scalable 3D generation.

Text-to-3D Generation with Bidirectional Diffusion using both 2D and 3D priors paper page: Most 3D generation research focuses on up-projecting 2D foundation models into the 3D space, either by minimizing 2D Score Distillation Sampling (SDS) loss or fine-tuning on multi-view datasets. Without explicit 3D priors, these methods often lead to geometric anomalies and multi-view inconsistency. Recently, researchers have attempted to improve the genuineness of 3D objects by directly training on 3D datasets, albeit at the cost of low-quality texture generation due to the limited texture diversity in 3D datasets. To harness the advantages of both approaches, we propose Bidirectional Diffusion(BiDiff), a unified framework that incorporates both a 3D and a 2D diffusion process, to preserve both 3D fidelity and 2D texture richness, respectively. Moreover, as a simple combination may yield inconsistent generation results, we further bridge them with novel bidirectional guidance. In addition, our method can be used as an initialization of optimization-based models to further improve the quality of 3D model and efficiency of optimization, reducing the generation process from 3.4 hours to 20 minutes. Experimental results have shown that our model achieves high-quality, diverse, and scalable 3D generation.

AK

38,853 Aufrufe • vor 2 Jahren

DisCo: Disentangled Control for Referring Human Dance Generation in Real World paper page: Generative AI has made significant strides in computer vision, particularly in image/video synthesis conditioned on text descriptions. Despite the advancements, it remains challenging especially in the generation of human-centric content such as dance synthesis. Existing dance synthesis methods struggle with the gap between synthesized content and real-world dance scenarios. In this paper, we define a new problem setting: Referring Human Dance Generation, which focuses on real-world dance scenarios with three important properties: (i) Faithfulness: the synthesis should retain the appearance of both human subject foreground and background from the reference image, and precisely follow the target pose; (ii) Generalizability: the model should generalize to unseen human subjects, backgrounds, and poses; (iii) Compositionality: it should allow for composition of seen/unseen subjects, backgrounds, and poses from different sources. To address these challenges, we introduce a novel approach, DISCO, which includes a novel model architecture with disentangled control to improve the faithfulness and compositionality of dance synthesis, and an effective human attribute pre-training for better generalizability to unseen humans. Extensive qualitative and quantitative results demonstrate that DISCO can generate high-quality human dance images and videos with diverse appearances and flexible motions.

DisCo: Disentangled Control for Referring Human Dance Generation in Real World paper page: Generative AI has made significant strides in computer vision, particularly in image/video synthesis conditioned on text descriptions. Despite the advancements, it remains challenging especially in the generation of human-centric content such as dance synthesis. Existing dance synthesis methods struggle with the gap between synthesized content and real-world dance scenarios. In this paper, we define a new problem setting: Referring Human Dance Generation, which focuses on real-world dance scenarios with three important properties: (i) Faithfulness: the synthesis should retain the appearance of both human subject foreground and background from the reference image, and precisely follow the target pose; (ii) Generalizability: the model should generalize to unseen human subjects, backgrounds, and poses; (iii) Compositionality: it should allow for composition of seen/unseen subjects, backgrounds, and poses from different sources. To address these challenges, we introduce a novel approach, DISCO, which includes a novel model architecture with disentangled control to improve the faithfulness and compositionality of dance synthesis, and an effective human attribute pre-training for better generalizability to unseen humans. Extensive qualitative and quantitative results demonstrate that DISCO can generate high-quality human dance images and videos with diverse appearances and flexible motions.

AK

161,479 Aufrufe • vor 3 Jahren

Spatial reconstruction is a long-context problem: real scenes come with hundreds of images. But O(N²) transformer-based models don’t scale efficiently. Introducing: 🤐ZipMap (CVPR ’26): Linear-Time, Stateful 3D Reconstruction via Test-Time Training (TTT). ZipMap “zips” a large image collection into an implicit TTT scene state in a single linear-time operation. The state will then be decoded into spatial outputs, and can be queried efficiently for novel-view geometry and appearance (~100 FPS) ZipMap is not only much faster (>20× faster than VGGT), but also matches or surpasses the accuracy of all SOTA models.

Spatial reconstruction is a long-context problem: real scenes come with hundreds of images. But O(N²) transformer-based models don’t scale efficiently. Introducing: 🤐ZipMap (CVPR ’26): Linear-Time, Stateful 3D Reconstruction via Test-Time Training (TTT). ZipMap “zips” a large image collection into an implicit TTT scene state in a single linear-time operation. The state will then be decoded into spatial outputs, and can be queried efficiently for novel-view geometry and appearance (~100 FPS) ZipMap is not only much faster (>20× faster than VGGT), but also matches or surpasses the accuracy of all SOTA models.

Haian Jin

79,916 Aufrufe • vor 6 Monaten

HunyuanWorld-Voyager is here and fully open-source! The world’s first ultra-long-range world model with native 3D reconstruction, redefining AI-driven spatial intelligence for VR, gaming, and simulations. ✅Direct 3D Output: Exports point cloud videos to 3D formats without tools like COLMAP, enabling instant 3D application use. ✅Innovative 3D Memory: Introduces a scalable world caching mechanism, ensuring geometric consistency across any camera trajectory. ✅Top-Ranked Performance: #1 on Stanford’s WorldScore, excelling in video generation and 3D reconstruction benchmarks.( Built on HunyuanWorld 1.0, Voyager blends video generation with 3D modeling, delivering camera-controlled, high-fidelity RGB-D sequences. Control scenes via keyboard or joystick for unmatched 3D consistency. Explore now: 🌐Project Page: 🔗GitHub: 🤗HuggingFace: 📝Technical Details:

HunyuanWorld-Voyager is here and fully open-source! The world’s first ultra-long-range world model with native 3D reconstruction, redefining AI-driven spatial intelligence for VR, gaming, and simulations. ✅Direct 3D Output: Exports point cloud videos to 3D formats without tools like COLMAP, enabling instant 3D application use. ✅Innovative 3D Memory: Introduces a scalable world caching mechanism, ensuring geometric consistency across any camera trajectory. ✅Top-Ranked Performance: #1 on Stanford’s WorldScore, excelling in video generation and 3D reconstruction benchmarks.( Built on HunyuanWorld 1.0, Voyager blends video generation with 3D modeling, delivering camera-controlled, high-fidelity RGB-D sequences. Control scenes via keyboard or joystick for unmatched 3D consistency. Explore now: 🌐Project Page: 🔗GitHub: 🤗HuggingFace: 📝Technical Details:

Tencent Hy

198,391 Aufrufe • vor 1 Jahr

3D-LLM: Injecting the 3D World into Large Language Models paper page: Large language models (LLMs) and Vision-Language Models (VLMs) have been proven to excel at multiple tasks, such as commonsense reasoning. Powerful as these models can be, they are not grounded in the 3D physical world, which involves richer concepts such as spatial relationships, affordances, physics, layout, and so on. In this work, we propose to inject the 3D world into large language models and introduce a whole new family of 3D-LLMs. Specifically, 3D-LLMs can take 3D point clouds and their features as input and perform a diverse set of 3D-related tasks, including captioning, dense captioning, 3D question answering, task decomposition, 3D grounding, 3D-assisted dialog, navigation, and so on. Using three types of prompting mechanisms that we design, we are able to collect over 300k 3D-language data covering these tasks. To efficiently train 3D-LLMs, we first utilize a 3D feature extractor that obtains 3D features from rendered multi- view images. Then, we use 2D VLMs as our backbones to train our 3D-LLMs. By introducing a 3D localization mechanism, 3D-LLMs can better capture 3D spatial information. Experiments on ScanQA show that our model outperforms state-of-the-art baselines by a large margin (e.g., the BLEU-1 score surpasses state-of-the-art score by 9%). Furthermore, experiments on our held-in datasets for 3D captioning, task composition, and 3D-assisted dialogue show that our model outperforms 2D VLMs. Qualitative examples also show that our model could perform more tasks beyond the scope of existing LLMs and VLMs.

3D-LLM: Injecting the 3D World into Large Language Models paper page: Large language models (LLMs) and Vision-Language Models (VLMs) have been proven to excel at multiple tasks, such as commonsense reasoning. Powerful as these models can be, they are not grounded in the 3D physical world, which involves richer concepts such as spatial relationships, affordances, physics, layout, and so on. In this work, we propose to inject the 3D world into large language models and introduce a whole new family of 3D-LLMs. Specifically, 3D-LLMs can take 3D point clouds and their features as input and perform a diverse set of 3D-related tasks, including captioning, dense captioning, 3D question answering, task decomposition, 3D grounding, 3D-assisted dialog, navigation, and so on. Using three types of prompting mechanisms that we design, we are able to collect over 300k 3D-language data covering these tasks. To efficiently train 3D-LLMs, we first utilize a 3D feature extractor that obtains 3D features from rendered multi- view images. Then, we use 2D VLMs as our backbones to train our 3D-LLMs. By introducing a 3D localization mechanism, 3D-LLMs can better capture 3D spatial information. Experiments on ScanQA show that our model outperforms state-of-the-art baselines by a large margin (e.g., the BLEU-1 score surpasses state-of-the-art score by 9%). Furthermore, experiments on our held-in datasets for 3D captioning, task composition, and 3D-assisted dialogue show that our model outperforms 2D VLMs. Qualitative examples also show that our model could perform more tasks beyond the scope of existing LLMs and VLMs.

AK

249,798 Aufrufe • vor 3 Jahren

4DGT: Learning a 4D Gaussian Transformer Using Real-World Monocular Videos Abstract: We propose 4DGT, a 4D Gaussian-based Transformer model for dynamic scene reconstruction, trained entirely on real-world monocular posed videos. Using 4D Gaussian as an inductive bias, 4DGT unifies static and dynamic components, enabling the modeling of complex, time-varying environments with varying object lifespans. We introduced a novel density control strategy in training, which allows our 4DGT to handle longer space-time input while maintaining efficient rendering at runtime. Our model processes 64 consecutive posed frames in a rolling-window fashion, predicting consistent 4D Gaussians in the scene. Unlike optimization-based methods, 4DGT performs purely feed-forward inference, reducing reconstruction time from hours to seconds and scaling effectively to long video sequences. Trained only on large-scale monocular posed video datasets, 4DGT can significantly outperform prior Gaussian-based networks in real-world videos and achieve on-par accuracy with optimization-based methods on cross-domain videos.

4DGT: Learning a 4D Gaussian Transformer Using Real-World Monocular Videos Abstract: We propose 4DGT, a 4D Gaussian-based Transformer model for dynamic scene reconstruction, trained entirely on real-world monocular posed videos. Using 4D Gaussian as an inductive bias, 4DGT unifies static and dynamic components, enabling the modeling of complex, time-varying environments with varying object lifespans. We introduced a novel density control strategy in training, which allows our 4DGT to handle longer space-time input while maintaining efficient rendering at runtime. Our model processes 64 consecutive posed frames in a rolling-window fashion, predicting consistent 4D Gaussians in the scene. Unlike optimization-based methods, 4DGT performs purely feed-forward inference, reducing reconstruction time from hours to seconds and scaling effectively to long video sequences. Trained only on large-scale monocular posed video datasets, 4DGT can significantly outperform prior Gaussian-based networks in real-world videos and achieve on-par accuracy with optimization-based methods on cross-domain videos.

MrNeRF

34,782 Aufrufe • vor 1 Jahr

Stable Virtual Camera: Generative View Synthesis with Diffusion Models Hint: Check the project website. It is awesome! Contributions: 1. A training strategy for jointly modeling large viewpoint changes and temporal smoothness. 2. A two-pass procedural sampling method for smooth video generation along arbitrarily long camera trajectories. 3. A comprehensive benchmark that evaluates NVS methods across different datasets and settings. 4. An open-source release of model weights to support future research.

Stable Virtual Camera: Generative View Synthesis with Diffusion Models Hint: Check the project website. It is awesome! Contributions: 1. A training strategy for jointly modeling large viewpoint changes and temporal smoothness. 2. A two-pass procedural sampling method for smooth video generation along arbitrarily long camera trajectories. 3. A comprehensive benchmark that evaluates NVS methods across different datasets and settings. 4. An open-source release of model weights to support future research.

MrNeRF

38,746 Aufrufe • vor 1 Jahr

Blended-NeRF: Zero-Shot Object Generation and Blending in Existing Neural Radiance Fields paper page: Editing a local region or a specific object in a 3D scene represented by a NeRF is challenging, mainly due to the implicit nature of the scene representation. Consistently blending a new realistic object into the scene adds an additional level of difficulty. We present Blended-NeRF, a robust and flexible framework for editing a specific region of interest in an existing NeRF scene, based on text prompts or image patches, along with a 3D ROI box. Our method leverages a pretrained language-image model to steer the synthesis towards a user-provided text prompt or image patch, along with a 3D MLP model initialized on an existing NeRF scene to generate the object and blend it into a specified region in the original scene. We allow local editing by localizing a 3D ROI box in the input scene, and seamlessly blend the content synthesized inside the ROI with the existing scene using a novel volumetric blending technique. To obtain natural looking and view-consistent results, we leverage existing and new geometric priors and 3D augmentations for improving the visual fidelity of the final result. We test our framework both qualitatively and quantitatively on a variety of real 3D scenes and text prompts, demonstrating realistic multi-view consistent results with much flexibility and diversity compared to the baselines. Finally, we show the applicability of our framework for several 3D editing applications, including adding new objects to a scene, removing/replacing/altering existing objects, and texture conversion.

Blended-NeRF: Zero-Shot Object Generation and Blending in Existing Neural Radiance Fields paper page: Editing a local region or a specific object in a 3D scene represented by a NeRF is challenging, mainly due to the implicit nature of the scene representation. Consistently blending a new realistic object into the scene adds an additional level of difficulty. We present Blended-NeRF, a robust and flexible framework for editing a specific region of interest in an existing NeRF scene, based on text prompts or image patches, along with a 3D ROI box. Our method leverages a pretrained language-image model to steer the synthesis towards a user-provided text prompt or image patch, along with a 3D MLP model initialized on an existing NeRF scene to generate the object and blend it into a specified region in the original scene. We allow local editing by localizing a 3D ROI box in the input scene, and seamlessly blend the content synthesized inside the ROI with the existing scene using a novel volumetric blending technique. To obtain natural looking and view-consistent results, we leverage existing and new geometric priors and 3D augmentations for improving the visual fidelity of the final result. We test our framework both qualitatively and quantitatively on a variety of real 3D scenes and text prompts, demonstrating realistic multi-view consistent results with much flexibility and diversity compared to the baselines. Finally, we show the applicability of our framework for several 3D editing applications, including adding new objects to a scene, removing/replacing/altering existing objects, and texture conversion.

AK

62,768 Aufrufe • vor 3 Jahren